27 days of AI bots on a live store

Measured by RobotsLLM itself on akumulyator.center, 16-29 August 2026. Catalog of about 4,600 products. This is load transparency, not a sales promise.

Stream of AI bots hitting an online store server

Method

Period: 16-29 August 2026 (14 days). Store: akumulyator.center, a batteries and solar shop with about 4,600 products.

What was measured: requests from a registry of 31 bots, verification against provider IP ranges or rDNS (Googlebot, Bingbot), AI referrals via UTM and HTTP Referer (ChatGPT, Gemini, Perplexity), and orders attributed to those sessions.

Verification checks the request IP against the provider's official ranges and rDNS. It runs in nightly batches, so part of the volume is confirmed (10,047 requests at the time of this snapshot), not all of it. The rest stays marked unverified, which means not confirmed, not fake.

279 289

bot requests

IP and rDNS bot verification

58

visits from AI assistants

ChatGPT 76 · Gemini 19 · Perplexity 2

3

orders from those visits

UTM / referrer attribution

Top bots

Bot Requests Purpose
meta-externalagent114,199training
googlebot74,439search
bingbot36,935search
chatgpt-user17,808user
amazonbot12,043search
claudebot9,688training
applebot7,322search
perplexitybot3,041search
bytespider2,106training
ccbot826training

45% of all bot requests came from training bots, 114,199 of them from a single Meta bot. Without an access policy that is simply a server bill.

Findings (honest)

Three orders in 14 days is not commercial proof. The sample is small, AI-referral attribution is incomplete (not every assistant sends a referrer), and we do not claim that llms.txt "sold" those orders.

The proof is different: the load is visible, bots are verified, training can be told from user-triggered traffic, and Allow / Disallow is a decision with numbers, not a guess.

How to reproduce this on your store

  1. Install the zip as a regular CS-Cart add-on. Install does not publish robots.txt.
  2. Pick a preset (Growth by default), inspect the diff, and publish the policy yourself.
  3. Enable cron: queue, log parser, aggregation, nightly audit.
  4. After a few days the dashboard shows bots, verification, and referrals down to orders.

Download free